The Pima County Election Integrity Commission is holding a special meeting today to talk about whether one member's statistical analysis of votes in recent elections shows evidence of fraud.
Under the group's bylaws, an emergency meeting can be held if at least five of the nine members call for one.
Mickey Duniho, a retired National Security Agency computer programmer, requested the meeting two days ago after he began plotting cumulative votes by precinct and noticing that outcomes seemed to differ by precinct size.
He was replicating earlier studies done by California researchers Francois Choquette and James Johnson, an aerospace engineer and a financial analyst. The researchers argue that their analysis of the recent Republican primary shows Mitt Romney making strange vote gains in most states' large precincts.
Duniho - formerly a Republican election observer in Maryland, a supporter of Democrat-backed lawsuits against Pima County's Elections Department and now a registered independent - said that his results seem to parallel those of Choquette and Johnson, who tried to account for their findings using demographics.
He is now collecting demographic data by precinct to try to explain his results with other factors, such as whether a precinct is rural or the affluence of the precinct's residents.
Duniho suspects that the patterns he found show a 10 percent flip of votes in favor of the Republican candidate in the 2010 race between Raúl Grijalva and Ruth McClung and the race between Gabrielle Giffords and Jesse Kelly the same year, as well as votes switched to benefit Romney in the Republican primary.
"The problem is figuring out what the statistical evidence does mean," Duniho said. "The computer is a black box. It is very easy for the guy who wrote the program to do just about anything."
At today's meeting, Duniho hopes to persuade the county Elections Department to sort early ballots by precinct before doing the hand-count audit required by law.
He has been advocating for that sorting, as well as for upping the percentage of ballots hand-counted, for about six years, arguing that his method boosts the chances of revealing fraud if it were to occur.
By law, Arizona counties must do a hand-count audit of 1 percent of early ballots and 2 percent of precincts in at least one federal and one state race. Pima County already audits more than required - 4 percent of precinct-cast ballots and 1 percent of early ballots. No local races are audited.
Some of the commission's members have argued strongly against holding the meeting and worry that it could unnecesarily increase fears about the vote count.
Benny White, a Republican election observer, responded to news of the meeting request with a sharply worded email.
"After reviewing the academic research involved with the links in the message, I conclude that the allegations being made are absolute nonsense," he wrote. "These academics don't take into account the fact that election results are the response by voters to campaigns and candidates. …
"I think there is a greater probability that fluctuations in the electrical voltage of the lines serving the election department have more to do with variations in election results than these alleged anomalies."
The county's technical consultant on election matters, John Moffatt, agrees that the data do not seem to show a vote flip in Pima County, but he does think the California researchers may be on to something with their findings in some other states.
"It's worth paying attention to, and we took it seriously," he said. "My personal opinion is that it's another witch hunt, but our responsibility is to check this stuff out, not just blow it off."
He adamantly rejects allegations that county elections staff somehow tampered with any results.
The county's elections director, Brad Nelson, will not be at the meeting to approve a change of audit procedures because of family issues, but county workers involved in those processes caution that while it's theoretically possible to make Duniho's suggested change, it would be logistically difficult.
"That's a monumental task," Pima County Recorder F. Ann Rodriguez said. "It's kind of late to be changing the procedures in the middle of a major election."
The sorting machine needed to do the job efficiently would cost at least $125,000, said Chris Roads, deputy recorder and registrar of voters.
To do the sorting by hand would likely take two days, Moffatt said.
The window to challenge a vote count after an election in Arizona is five days after the canvass.
On StarNet: Go to azstarnet.com/news/local/govt-and-politics to read more about local and state government and political news.
Contact reporter Carli Brosseau at email@example.com or 573-4197. On Twitter @carlibrosseau